AI CV Screening Scorecard Template: What to Include Before You Upload CVs

An AI CV screening scorecard is the agreed set of hiring criteria you use before any candidate is reviewed. It tells the screening tool what matters, how much each requirement matters, what evidence counts, and when a human should take a closer look.

That matters because most screening problems start before the first CV is opened. The job description says one thing, the hiring manager cares about something slightly different, and the recruiter is left to infer the real standard across 100 or 200 applications.

A good scorecard fixes that. It turns a vague role brief into a consistent review process.

The Quick Template

Use this structure before uploading CVs to any screening tool:

Section	What to define	Example
Must-have criteria	Requirements a candidate needs for serious consideration	Right to work in the UK, 3+ years in B2B sales
Weighted criteria	Skills or experience that should affect ranking	30% sector experience, 25% quota ownership, 20% CRM depth
Nice-to-have signals	Evidence that improves confidence but should not exclude someone	Enterprise deal cycles, HubSpot experience, partner sales
Red flags	Issues that need human review	Unclear dates, unexplained seniority gaps, location limits
Evidence rules	What counts as proof	Named project, measurable result, repeated responsibility
Review bucket	The action that should follow	Interview, review manually, ask follow-up, reject

If your scorecard cannot fit on one page, it is probably too complicated for first-pass screening.

Start With the Hiring Decision

Before listing skills, write down the actual decision you need to make.

For most first-pass CV screening, the decision is not "Who should we hire?" It is:

Which candidates deserve human attention first?

That distinction keeps the scorecard honest. First-pass screening should help you shortlist, hold, or reject with reasons. It should not pretend to replace interviews, work samples, references, or hiring manager judgement.

Marxel uses this idea directly: candidates are sorted into action-based buckets such as aligned, potential, hold, and unclear. You can see how this fits the wider workflow in how Marxel works.

Separate Must-Haves From Ranking Factors

One common mistake is treating every requirement as equally important. That turns screening into keyword bingo.

Instead, split criteria into two groups:

Must-haves

These are the minimum conditions for the role. Missing one should usually push a candidate into manual review or rejection.

Examples:

• Required certification or licence
• Right to work or location requirement
• Minimum seniority for a regulated role
• Specific language requirement
• Essential technical environment

Keep this list short. If everything is a must-have, nothing is.

Ranking factors

These help distinguish strong candidates from acceptable ones.

Examples:

• Similar industry experience
• Scale of team or budget owned
• Depth in a specific tool
• Evidence of repeated progression
• Measurable outcomes

Ranking factors should have weights. A candidate with weaker tool overlap but stronger industry experience may still be worth interviewing if the scorecard reflects that tradeoff.

Define What Evidence Counts

AI screening becomes unreliable when the criteria are clear but the evidence rules are vague.

Bad criterion:

Strong stakeholder management.

Better criterion:

Evidence of managing cross-functional stakeholders, shown through project ownership, client delivery, steering groups, or explicit collaboration with product, engineering, finance, legal, or operations teams.

This helps the model avoid shallow matching. It also helps humans review the output because each decision can point back to the same standard.

For compliance-sensitive roles, this is especially important. Candidate decisions should be explainable in plain English, not hidden behind a score. Our GDPR-compliant CV screening guide covers the legal side of this in more detail.

Add Red Flags Without Automating Rejection

Red flags should usually trigger human review, not automatic rejection.

Useful red flags include:

• Missing work dates
• Unclear employment status
• Claims that need verification
• Location ambiguity
• Very short tenure patterns
• Required credential not visible

The safe pattern is: flag the concern, explain the evidence, and let a person decide.

That is different from using the AI tool as a rejection machine. A red flag should create a question, not end the process by itself.

Use Four Action Buckets

For first-pass screening, four buckets are usually enough:

Bucket	Meaning	Next step
Aligned	Strong fit against must-haves and priorities	Send to hiring manager or interview stage
Potential	Some strong evidence, but not a clean match	Human review before deciding
Hold	Fit depends on missing or unclear detail	Ask follow-up question or check evidence
Unclear	Missing important requirements	Reject after human confirmation if needed

This is easier to act on than a 0-100 score. A hiring team does not need a false sense of precision. It needs the next step.

Example Scorecard

Here is a simplified scorecard for a Customer Success Manager role:

Criterion	Type	Weight	Evidence rule
B2B SaaS customer success	Must-have	-	Named CS, account management, or implementation role
Owns renewals or expansion	Weighted	25%	Renewal targets, expansion work, commercial ownership
Handles 50+ customer accounts	Weighted	20%	Portfolio size, segment ownership, book of business
Works with product or support	Weighted	15%	Escalations, feedback loops, technical coordination
CRM and customer health tooling	Nice-to-have	10%	Gainsight, HubSpot, Salesforce, Planhat, Vitally
Enterprise stakeholder work	Weighted	20%	Senior contacts, QBRs, procurement, legal cycles
Unclear employment dates	Red flag	-	Missing dates or overlapping roles

Before processing CVs, ask the hiring manager to confirm:

• Which criteria are genuinely non-negotiable?
• Which tradeoffs are acceptable?
• Which red flags should create a follow-up question?
• What would make a surprising candidate worth reviewing?

How to Use This With AI Screening

The workflow should look like this:

1. Upload the job description.
2. Add hiring manager notes.
3. Generate or draft the scorecard.
4. Review the scorecard before candidate processing.
5. Screen CVs against the confirmed criteria.
6. Review buckets and reasoning.
7. Move candidates forward only after human confirmation.

If a tool skips step 4, be careful. You want control before candidate data is processed, not after the model has already made assumptions.

Related Reading

• Manual vs automated CV screening
• Best AI CV screening software UK
• How long does it take to screen 100 resumes?
• CV screening calculator

---

Want to test a scorecard on a real role? Upload a job description to Marxel, review the generated rubric, and screen up to 25 CVs free. Start free screening